Assessing the spatial extent of dryland salinity through fuzzy modeling

This paper presents applied research on fuzzy logic modeling to predict the distribution of secondary dryland salinity. An existing approach to predicting the distribution of salinity, fuzzy landscape analysis GIS (FLAG), developed by Roberts et al. (1997) is implemented. An attempt is made to optimize the predictive power of FLAG through the inclusion of geological and vegetation data. As FLAG models salinity distribution within the framework of fuzzy logic, results from this investigation are compared with the outputs of a predictive model of salinity based on probability theory. tempt to optimize FLAG was not as successful as expected. Of the FLAG based predictions, the model derived from fuzzy discharge indices (CC) produced the most accurate result. Of the modified FLAG models, FLAG_VEG (that incorporates vegetation data) produced the best result. The prediction of areas at risk of salinity derived from the probabilistic model presented similar accuracy when validated against ground truth data used to validate the results of this research. The comparison of CC and FLAG_VEG with the probability-based prediction model of salinity indicated that the differences between them were not significant at a 95% confidence level, with the fuzzy logic based models outperforming the probabilistic model in steeper terrain.